[lfs-dev] Future of i686 builds (in gcc, et al.)

Paul Rogers paulgrogers at fastmail.fm
Wed Mar 7 11:57:20 PST 2018

I recently had to put a 1GHz Pentium-III "Coppermine" in service.  One of several things, it is for testing my previous LFS-7.10 build.  I built all 341 packages in two systems, i686 and x86-64.  The i686 version runs fine on my Core2-Duo systems, but I wanted to check that it runs as advertized in a "real" i686 with no extra instructions on the side.

In my definition of "general operations" it works well enough, but don't run Libre-Office.  I had patched the kernel up to LTS 4.1.42, and had to rebuild it for the i810 hardware.  Duration: just short of 2 hours!

I call that a "heavy lift".  So my question morphed into: when does i686 hardware just lack the horsepower to "run" modern, i.e. kernel-3.x or kernel-4.x, LFS systems?  On what does it make sense to abandon i686?  Most public distros did that some time ago, but obviously I keep lots of old hardware around with some attempt at functionality.  I have a functional 2GHz Pentium-4 "Northwood" box, so I copied the drive for the Coppermine box and slipped it in.  (The CPU flags include HT, but trying an SMP/HT kernel can't find it.)

Compiling LFS-7.10 linux-4.1.42 duration: ~52 minutes.

Other systems on the drive/box:

Compiling LFS-7.7 linux-3.10.107 duration: ~22 minutes.
Compiling LFS-7.2 linux- duration: ~18 minutes.

So, as I now work through LFS-8.1, and read the options in building gcc and other places, it seems to me kernel 4.x-based software has just gotten too "heavy" for i686 to be a reasonable build.

I leave it up to LFS-devs to decide when to take it out of the book.  As much as I regret it, it's probably about time, so I won't.

p.s. I still hope to get Meltdown/Spectre patches for linux-4.1.x i686 builds.

Paul Rogers
paulgrogers at fastmail.fm
Rogers' Second Law: "Everything you do communicates."
(I do not personally endorse any additions after this line. TANSTAAFL :-)

More information about the lfs-dev mailing list